Toyota Research Institute (TRI) has revealed new robotics capabilities to solve complex tasks in the home environment. Roboticists were able to train robots to understand and operate in complicated situations that would confuse typical robots. The complex situations included recognizing and responding to transparent and reflective surfaces in a variety of circumstances.
TRI demonstrated these new capabilities in a video, which can be seen below, highlighting a system allowing robots to make generalizations in various scenarios, including different homes. The video has a humorous tone and shows the Toyota robot capturing its new capabilities on film that was released on National Selfie Day as it performed its newly learned skills around the home. TRI VP of robotics Max Bajracharya says training robots to operate in home environments has its own set of special challenges because of the diversity and complexity of homes.
That complexity means that small tasks can become significant challenges for robots. In the home environment, humans can easily differentiate between an object and its reflection. However, transparent reflective items often found in homes today can confuse typical robots. Most robots are programmed to react to the objects and geometry in front of them without considering the context of the situation. They can be confused by things such as glass tables, reflective appliances, or transparent cups.
TRI developed a new training method to perceive the 3D geometry of the scene while detecting objects and surfaces. The combination allowed researchers to use large amounts of synthetic data to train their system. The use of synthetic data alleviated the need for time-consuming, expensive, and impractical data collection and labeling.
TRI admits that no system is perfect, but today’s announcement adds to the body of knowledge that will help robots to navigate and operate in the home environment reliably. The advancement could help someday lead to robots to help with everyday tasks in the home for those who cannot perform the task themselves.